Population Health Management, Predictive Analytics, Big Data and Text Analytics

Population Health Management Analytics
3HCAD 6635 Health Information Analytics Copyright © 2016 Frank F. Wang 1

Anatomy of Healthcare Delivery
HCAD 6635 Health Information Analytics Copyright © 2016 Frank F. Wang 2

6
Population Health Management: Process and Business Requirements
• Stages to achieving a Value-based Accountable Care
Patient Panel
Definition
Targeted
Populations &
Outcomes
Baseline
Expenditures
& Costs
Accountability
Models
Financial
Reconciliation
Population Health
Management
• Identify Unique
Patients
• Assemble
Records of
Clinical Care
• Define Bundles
• Identify Unique
Providers
• Align Patients &
Providers
• Measure /
Manage Care
Delivery
• Measure /
Manage Care
Relationships
• Patient Panel
Analytics
• Defined Patients,
Beneficiaries or
Members
• Segmentation
• Outcomes:
Clinical,
Operational,
Financial
• Identify ACO
Parties & Roles
• Performance
Targets &
Metrics
• Targeted Care
Plans
• EBM Guidelines
for Required
Care for Patient
Needs
• Historical
Baselines
• Align Patient
with Provider
Entity
• Align Provider with
ACO Entity
• Calculate
Service Fees &
Savings Targets
• Hierarchical
Segmentation &
Aggregation
• Anticipated
Services,
Charges &
Costs
• Collaborative
Care Delivery
Models
• Transitions in
Care
• Communications,
Handoffs, Follow-
ups
• Contracts,
Roles,
Responsibilities
• Shared Metrics,
Benefits &
Risks
• Retrospective
Payments
• Shared Savings &
Costs
• Value Realization
• Allocated Gains
(Losses)
• Billing &
Payment
Distribution
• Compliance &
Adherence
Targets
• Patient
Stratification
• Comparative
Outcomes &
Quality Metrics
• Prospective &
Bundled
Payment Models
• Predictive Risk
Modeling
• Performance
Optimization
• Market Share &
Competitive
Analytics

7
Population Health Analytics: Analytic Needs Assessment
How do we manage patient cohorts systematically? How do we focus
and integrate our care delivery across populations & care settings?
Population Health
Management
Do we understand our charges, payments and costs? Are we
reconciling these with our care plans and our accountability models?
Financial
Reconciliation
How do we implement & measure accountability? Where and by
whom are value and costs introduced into our delivery processes?
Accountability
Models
What are our baseline data on these targets, with this payer? How do
these align with our contract terms across payer types?
Baseline Expenditures
& Costs
What are our current targets? What quality / results are we seeing?
Are they consistent? Where do we see under- or over-performance?
Targeted Populations
& Outcomes
Who are patients? What treatments are they receiving? What other
providers are they seeing? At what locations? With what frequency?
Patient Panel
Definition

24
Population Health Management Analytics for Value-Based Healthcare
PHC
Analytics
Clinical
Strategic
Planning
IT
Practice
Mgmt
Marketing
Finance
• Revenue Cycle
• Costs, Margin
• Payer Mix
• Stratification
• Outcomes
• Quality &
Safety
• Growth
• Market Share
• Competition
• Architecture
• Data Quality
• Tools, Applications
• Security, Governance
• Patient Satisfaction
• Panel Management
• Continuum of Care
• Outreach
• Physician Liaison
• Relationship Mgmt
• Service
Improvement
• Integrating Analytics for Clinical, Operational and Financial Improvement

The 8 Building Blocks of Successful Accountable HealthcarePayforReportingPayforOutcomes
EHR/PMS/
E-Prescribing 2. Automating and Integrating Fragmented Stakeholders
Information
Exchange
(HIE)
3. Sharing Clinical, Operations and Financial Information
Aggregation &
Analytics 4. Aggregating Siloed Data and Gaining Insight
Decision
Support 5. Transforming collected data into clinical knowledge
Healthcare
Portals and
Medical Homes
6. Making clinical information accessible and “team-based” care possible
Outcomes
Measurement &
Reporting
7. Establishing Core Measures and Reporting Outcomes
Risk
Sharing 8. Enabling Population Based Management and Risk Sharing Models
Converged
Medical
Infrastructure
1. Establishing Standardized and Optimized IT Platforms

Population Health Management: Challenges in Culture, Stakeholders,
Data, Technology and Governance
• Data From Multiple Source Systems of Record and Points of Origin
 Differing Formats and Semantics
 Inconsistent Taxonomies
 Differing Data Granularities
• Technical Challenges
 Timing and Granularity Differences and Conflicts
 Access to data stored in the cloud
 Positioning for Big Data Opportunities
• End-User Experience
 Consistent but Responsive (Variable, Tailorable) Experience
 Power User vs. Ease of Use
 Education on Source, Meaning and Veracity of Data Elements
• Data Governance
 Lack of consistent Enterprise-wide definitions
 Different groups use similar terminology for different data and meanings

11
Population Health Management: Navigating Complex Data Spaces
Patient
Organization
Provider
Location
Contracts
Payer
Claims
Payments
Encounter
Charges
Costs
Diagnosis
Treatments
Chronic
Condition
Disease
Group
Procedures
Medications
Margin
Events
Data Flow Model …

8
Population Health Management: Data Integration Challenges
• Demographics
• History
• Reported
Outcomes
• Location • Specialty
• Relationships
• Location
• Care Team
• Structure
• Locations
• Legal Entity
• Contracts
• Care Mgmt
Teams
• Inpatient
• Outpatient
• Pharmacy
• Beneficiary
History
• Payers
• Charges,
Payments &
Adjustment
• Costs
• Margin
• Risk
Contracts
• Diagnosis
• Chronic
Conditions
• Labs & Results
• Procedures &
Medications
• Quality
• Appts
Scheduling
• Utilization &
Throughput
• DRG
• Location

9
Population Health Management Enterprise Data Architecture
DataIntegration&Transformation
Dashboards &
Analytic Views
Contract Measures
Performance
Summary
Baseline Expenditure
Provider Profile
DataAccess–Navigation&Security
Reports
Capture Integration and Transformation Consumption
Extensible Data
Architecture
Provider
Standard Data Models
Patient
Location
Claim
Reference
Other
Master
Data
Encounter
Patient Panel Analytics
Targeted Populations
& Outcomes
Baseline Expenditures
& Costs
Accountability Models
Financial Reconciliation
Population
Health
Management
Health
System
EMR
Billing
MPI
Provider
Master
Coding
Payers
Members
Claims

23
Population Healthcare Management Analytics
Accountable Care and Population Management Platform
Patient Panel
Definition
Targeted
Populations &
Outcomes
Baseline
Expenditures
& Costs
Accountability
Models
Financial
Reconciliation
Population
Health
Management
Integrated
Data Platform
Changes to Processes
& Operations
Changing
Healthcare Model
Population &
Practice Models

22
Population Health Management: Mobilizing for Value-Based Care
Workflows managing integrate data, care delivery, communications and metrics
• Physician Office
• Other Care Settings
Labs
Data Capture
• Analytics
• Patient Registry
• Financial & Quality
Measures
Workflow Triggers,
Alerts & Escalation
• Patient Registration,
Scheduling
• Call Center
• Patient Home
• Web Access
• Progress
Review
• Assessment & Stratification
• Individualized Care Plan
• Discharge
EMRs
Quality
Performance
• Outreach
Workflow Management
Patient Engagement
Care
PlansPractice
Mgmt
Cost Models
Integration & HIE
Metrics

Predictive Analytics

• Predictive analytics is the practice of extracting information from
existing data sets in order to determine patterns and predict future
outcomes and trends.
• It does not tell you what will happen in the future.
 It forecasts what might happen in the future with an acceptable level of
reliability, and includes what-if scenarios and risk assessment.
• Gartner goes a step further:
 Analysis measured in hours or days (real-time or near real-time).
 The emphasis on the business relevance of the resulting insights, like
understanding the relationship between x and y.
 An emphasis on ease of use, thus making the tools accessible to business
users.
Source: www.gartner.com
What is Predictive Analytics?

Predictive Analytics in Population Health Management
• It must be
timely
• It must be
role-specific
• It must
actionable
Risk scores
(stratification)
What-if scenarios
(simulation)
Geo-spatial analysis
(mapping)
HCAD 6635 Health Information Analytics
Copyright © 2016 Frank F. Wang 15

• Risk stratification scoring can
assist in
• Prioritizing clinical
workflow
• Reducing system waste
• Creating financially
efficient population
management.
• Well-established risk
stratification scores of low-risk,
high-risk, and rising-risk can
play a key role in several
healthcare scenarios.
Risk Stratification

Overview of Risk Stratification Methods
• Hierarchical Condition Categories (HCCs) - CMS Medicare Advantage Program
 Contains 70 condition categories selected from ICD codes and includes expected health
expenditures.
• Adjusted Clinical Groups (ACG) - Johns Hopkins University
 Uses both inpatient and outpatient diagnoses to classify each patient into one of 93 ACG
categories. It is used to predict hospital utilization.
• Elder Risk Assessment (ERA)
 Uses age, gender, marital status, number of hospital days over the prior two years, and selected
comorbid medical illness to assign an index score to each patient (for adults 60 years and older).
• Chronic Comorbidity Count (CCC) - Clinical Classification Software from AHRQ
 Is the total sum of selected comorbid conditions grouped into six categories.
• Minnesota Tiering (MN) – Major Extended Diagnostic Groups (MEDCs)
 Groups patients into one of five tiers from Tier 0 (Low: 0 Conditions), Tier 1 (Basic: 1 to 3), Tier 2
(Intermediate: 4 to 6), Tier 3 (Extended: 7 to 9), to Tier 4 (Complex: 10+ Conditions).
• Charlson Comorbidity Measure - The Charlson Model
 Predicts the risk of one-year mortality for patients with a range of comorbid illnesses.
 Uses administrative data.
 Categorizes the presence/absence of 17 comorbidity definitions and assigns patients a score from
one to 20, with 20 being the more complex patients with multiple comorbid conditions.
 Used for predicting future poor outcomes.

Stratifying Population and Predicting Risks Based on Comorbidity
• A Histogram (Frequency Distribution) of a
Charlson Index Score for a population of
heart failure patients.
• The Frequency Distribution can help PHM
to gauge level of risk of the population.
• Charlson/Deyo Model based
on multiple inputs:
 # and types of
comorbidity
 Age
 An index score of risk
assigned to each patient
 Used as a general risk
stratifier
 Used as a mortality
predictor
 Can be incorporated in
many applications as a
filter to identify high risk
patients

Creating a Predictive Model of Heart Failure Readmission Risks
• Multiple clinical and demographic parameters based on historical data
 Comorbidity Index
 Others include race, gender, family history, vital signs, length of stay (LOS), propensity
and predisposition of patients to diseases, etc.
 Regression on historical data ran to the weight that should be assigned to each
parameter in the model.
 Those weights determine the impact each parameter has on the predicting readmission
• Applying the predictive model to flag high-risk patients and suggest pro-
active actions needed

Stratifying Population and Predicting Risks Based on Comorbidity

Simulation/What-if Scenarios
• Calculating the amount of opportunity
dollars to capture in healthcare financial
administration to reduce variation in a
specific clinical care process.
• Provides clinicians and administrators a
safe glimpse into “what if” simulation and
assess the likely outcomes of a given
combination of events in healthcare
delivery .
• Optimizing campaign budget allocation in
healthcare marketing.
• Allows payers to define premiums
effectively in healthcare insurance
exchange markets

22
Areas Showing Benefits Now
• Improved Patient Flow
• Reduced Readmissions
• Disease Outbreak Prediction
• Emergency Room Risks
HCAD 6635 Health Information Analytics Copyright © 2016 Frank F. Wang

23
• Enables prediction which resources will be needed at any given time
• Predicting patient flow versus patient tracking
• Reduces bottlenecks and wait times
 Especially in the emergency room
 Increases patient satisfaction
Improve Patient Flow
• Improves admissions and discharges workflows
 Result in efficient patient placement at admission
 Find bottlenecks and drive for earlier or later discharge times
• Manages capacity needs
 Identify underused beds and labs to better target patient usage
 Improves patient care and increased revenues
• Transport and housekeeping
 Track job times and responsiveness to improve turnover

24
• Predict the risk of readmission in 30 days to a patient to assist
with the decision to release a patient
• Reduce cost of readmission and the opportunity cost of a patient
occupying a bed that could be used by others
• Requires a proactive versus reactive approach to be effective
Reduced Readmissions
• Identify the factors effecting readmissions
 Discover and use descriptive analytics based on prior readmission
data
 Create an algorithm to predict who is likely readmitted
• Create automated processes to identify patients who are at risk
for readmission based on clinical, demographics, etc.
 Counter with a strategic response
 Gain information immediately from failures
• Make sure personnel adhere to the identified strategy
 Periodically valuate effectiveness

Disease Surveillance Monitoring and Reacting to Outbreaks like Ebola
Monitoring chief complaint /reason for
admission data in Admit, Discharge,
and Transfer (ADT) data streams.
Monitoring coded data
collected in Electronic
Health Records (EHRs).
Monitoring
billing data.
• Diseases also have a data profile (symptoms, perhaps discrete lab
results or other diagnostics like imaging).
• Boolean logical determinations, based on complete and valid data,
may point to opportunities for computer-assisted treatment decision-
making.
Data Profile
Alert

Disease Surveillance in the Near-Term
• Monitoring ADT messages: Use the chief complaint/reason for
admission data in an ADT message.
Advantage: Real-time, upon presentation
of the patient at a healthcare facility.
Disadvantage: Lacks codified, computable
data in the data stream, requiring natural
language processing (NPL).
• Analyzing Coded EHR and Other Clinical Data: Monitors coded
data (SNOMED or ICD) for diagnosis, labs tests and results, and
diagnostic imaging.
Advantage: The most precise method;
unlikely to ever be a real-time option due to
the inherent nature of healthcare delivery.
Disadvantage: Timeliness of treatment data
will lag the decision making process too
late for effective decision making.
• Analyzing Coded Data From Billing Systems: This has all the
problems of the other two. It’s not unusual for revenue cycle
processes and systems to take over 30 days to drop a bill.

• Google Flu Trends shown to foresee an increase in influenza cases 7
to 10 days earlier than the CDC
 Descriptively analyzed online search trends
 Discovered that many search queries tend to be popular exactly
when flu season is happening
 Hypothesized people with flu symptoms seek information
 Compare query counts with traditional flu surveillance systems
 Identify correlation between how many people search for flu-related
topics and how many people actually have flu symptoms
 Aggregate all flu-related search queries to establish a pattern
 Estimate how much flu is circulating in different countries and
regions (Geo-special analytics)
 Can even pinpoint disease increase down to the hospital level
 Resources can be allocated to prepare for influx of patients
 Will discuss how to supplement this with text analytics
Disease Outbreak Prediction using Social Media Analytics

28
• Used to predict whether a patient is likely to:
 Go into cardiac arrest
 Suffer a stroke
 Potentially suffer from sepsis shock
• Collecting real time data (real-time operational analytics)
along with patient’s clinical history
 Compare to prior patient data
Emergency Room Use

Source: HealthIT Analytics, "Just 15% of Hospitals Use Predictive Analytics Infrastructure," Jennifer Bresnick, March 24, 2015
http://healthitanalytics.com/news/just-15-of-hospitals-use-predictive-analytics-infrastructure/
• Just 15% of hospitals are using advanced
predictive analytics to stay one step ahead
of preventable hospital readmissions, and
hospital-acquired conditions.
• Data-Driven: letting the data dictate how we treat and manage
customers with as much automated decisions as possible
• Data-Enabled: providers are provided real- time or near real-time information
to enable them to make better decisions and diagnoses based on hundreds
or thousands of patients with similar symptoms and demographics
Predictive Analytics is Key to Transform an Enterprise’s Culture
from Data-Driven to Data-Enabled

Data-Enabled Healthcare Organization Change Model: People,
Organization, Culture, Process, Data and Technology)
Paper-Based
Healthcare
Organization
Data-Enabled
Healthcare
Organization
Resistance
To Change
Isolated
Acceptance
Phase 3
Phase 4
DATA / TECHNOLOGY
ORGANIZATIONAL
/ PEOPLE
PROCESS /
WORKFLOWSMinimal
Data
Capture
Network-Wide
And Outside
Data Capture
Phase 3
Phase 4
Phase 4
EHR
Implementation
Analysis &
Modeling
Integration of
Data Sources
Predictive
30PrescriptiveHCAD 6635 Health Information Analytics Copyright © 2016 Frank F. Wang

Advanced Analytics

Big Data and Analytics
Big data “describes large volumes of high velocity, complex, and variable data
that require advanced techniques and technologies to enable the capture,
storage, distribution, management, and analysis of the information.”1
Big Data represents big opportunity
U.S. health care data alone reached 150 exabytes in 2011.
Big data for U.S. health care will soon reach zettabyte (1021 gigabytes) scale and
even yottabytes (1024 gigabytes) not long after.
32
1. Hartzband, D. D. (2011). Using Ultra-Large Data Sets in Health Care. 2011 Sessions (p. 3). e-healthpolicy.org.

Where Does Big Data Come From?
Web and social media data: Clickstream and interaction data from social
media such as Facebook, Twitter, Linkedin, and blogs.
Machine-to-machine data: Readings from sensors, meters, and other devices.
Transaction data: Health care claims and other billing records.
Biometric data: Fingerprints, genetics, handwriting, blood pressure, medical
images, retinal scans, and similar types of data.
Human-generated data: Unstructured and semi-structured data such as
electronic medical records (EMRs), physicians’ notes, email, and paper
documents.
33
SOURCE: Institute for Health Technology Transformation.

A Changing Healthcare Data Environment
34HCAD 6635 Health Information Analytics Copyright © 2016 Frank F. Wang
• Healthcare is entering into a phase of ‘post EMR’ deployment
where HCOs are “keen on gaining insights and instituting
organizational change from the vast amounts of data being
collected from their EMR systems”.
• HCOs must reduce costs and improve quality of care by
“applying advanced analytics to both internally and externally
generated data”.
• Larger volumes of structured and unstructured data can now be
managed and analyzed through “faster, more efficient and
cheaper computing (processors, storage, and advanced
software) and through pervasive computing (telecomputing,
mobile devices and sensors)”.
SOURCE: HIMSS, “What Is Big Data?”
http://www.himss.org/ResourceLibrary/genResourceFAQ.aspx?ItemNumber=30730

What Makes Big Data Big?
Factor Name Symbol # Bytes
103 kilobyte KB 1,024
106 megabyte MB 1,048,576
109 gigabyte GB 1,073,741,824
1012 terabyte TB 1,099,511,627,776
1015 petabyte PB 1,125,899,906,842,624
1018 exabyte EB 1,152,921,504,606,846,976
1021 zettabyte ZB 1,180,591,620,717,411,303,424
1024 yottabyte YB 1,208,925,819,614,629,174,706,176
35

The Four Dimensions of Healthcare Data
Adapted from:
Sun J. & Reddy CK. Big DataAnalytics for Healthcare.
http://dmkd.cs.wayne.edu/TUTORIAL/Healthcare
Big
Data
Volume
Veracity
VarietyVelocity

The Four Dimensions of Healthcare Data
• Volume – The amount of data being generated and stored. Typical
“big data” datasets can range from terabytes (1012 bytes) to
petabytes (1015 bytes) and exabytes (1018 bytes).
 Traditional database technologies (such as Relational Database
Management Systems, or RDBMS) and query tools (such as SQL)
are unable to scale efficiently to such volumes, necessitating new
approaches to data storage, management, and analysis.
• Variety – The number of different data sources has grown, ranging
from more traditional EMR data to website clickstream data and data
from social media sites (i.e., Twitter).
• Velocity – Refers to the speed at which data is generated
(through its numerous sources), accumulated (in associated
storage systems), and must be processed.
• Veracity – Not part of the original Gartner “big data” definition, but is
an indication of data quality (i.e., accuracy and completeness), trust
(credibility of the source), uncertainty, and suitability (of data for
target audience).

Big Data Classification
• http://www.ibm.com/developerworks/library/bd-archpatterns1

Big Data in Healthcare Today: Reality Check
• Volume: EMRs collect huge amounts of data,
but only half of the tables in an EMR database
(400 to 600 tables out of 1000s) are relevant to
the current practice of medicine and its
corresponding analytics use cases.
• Variety: Yes, but most systems collect very
similar data objects with an occasional tweak to
the model.
• Most health systems meet majority of analytics
and reporting needs today without big data.
• Not close to stretch the limits of what analytics
can accomplish with traditional relational
databases.
• Most healthcare institutions swamped with
some very pedestrian problems such as
regulatory reporting and operational
dashboards.

However, Big Data Is Coming To Healthcare
• New use cases (e.g. wearable medical
devices and sensors) drive the need
for big-data-style solutions.
• Embark on the journey of analyzing
text-based notes (chief complaints,
clinical charts).
• Big data indexing techniques add real
value to healthcare analytics.
• The introduction of genomic data and
precision medicine practices.
• Genomic sequences are huge files
and the analysis of genomes
generates even more data.

Big Data: the Internet of Things, Care Management and Clinical Trial
• SAS describes the IoT as:
a growing network of everyday objects from industrial machines to consumer
goods that can share information and complete tasks while you are busy
with other activities, like work, sleep, or exercise.
Volume, Variety, Velocity and Veracity:
• Wearable fitness devices collect personal health
data (heart rate, weight, trending) and sends that
data into the cloud will be part of this IoT.
• Accountable care organizations and Population
Health Management practices want to keep people
at home and out of the hospital will deploy sensors
and wearables.
• Healthcare institutions and care managers, using
sophisticated tools, will monitor this massive data
stream and the IoT to keep their patients healthy.
• Clinical trials of pharmaceuticals and medical
devices will deploy sensors to report outcomes.

Big Data: Predictive Analytics, Social economic Analytics and Geo-
spatial Analytics
• Real-time alerting.
• Predictive analytics used to dissect
socioeconomic data.
• Geo-spatial analytics used to dissect
geographic data
 Mapping layers and predictive analytics are
routinely used to forecast weather, optimize supply
chains, and support military deployment.
• Data show people in a certain zip code are
unlikely to have a car. A patient from that zip
code discharged from the hospital might have
difficulty going to a follow-up appointment at a
distant physician’s office
• Used to predict missed appointments and
noncompliance with medications.
 Visual and effective approach to decision-making.

Big Data: Predictive Analytics and Prescriptive Analytics
• Leverage historical data from other
patients with similar conditions, predictive
analytics can predict the trajectory of a
patient over time.
 Based on predictive algorithms
 Using programming languages such as
R and big data machine learning libraries
• Once we can accurately predict patient
trajectories, we can shift to the Holy Grail–
Prescriptive Analytics.
• Intervene to interrupt the patient’s
trajectory and set the proper course.
• Big data is well suited for these futuristic
use cases.

The Human Genome Project and Genomic Medicine
• The Human Genome Project, completed in
April 2003, made reading the full genetic
blueprint for human beings a reality.
• In the future clinicians will be able to
practice genomic medicine and
personalized care.
• It also has profound implications for the
future of analytics.
• The cost of sequencing a human genome
has fallen from about $ 1 billion in 2001 to
less than $1,000 today.
• As costs drop, advances in genomic
research are accelerating.

Using Genomics to Diagnose and Treat Disease
• A study published in the February
2014 edition of the NEJM
demonstrates that analyzing fetal
DNA in a pregnant woman’s blood
was a more accurate and less
invasive way of screening for
Down syndrome and other
chromosomal disorders than
methods such as ultrasound
imaging and blood tests.
• Discovery of the genomic defects for
more than 5,000 inherited diseases.
• Early genomic medicine success
lies in rare inherited diseases. These
diseases afflict more than 25 million
Americans.
• Many molecular diagnostic kits and
orphan drugs approved by FDA.

Using Genomics to Diagnose and Treat Disease
• The analysis of genomes is guiding
treatment for various types of cancer.
• Many cancer types can be
categorized by genomic traits and
divided into subtypes. Treatment are
being developed based on the
underlying genetic signature.
• This approach offers patients the
most efficacious treatment with
minimal side effects.
• Genomics is starting to be used to
improve the efficacy of medications
and how clinical care is delivered in
oncology, immunology and other
specialized medical practices.

Personalized Medicine through Genomics and Sensor Devices
• In the future, physicians will tailor treatment for
many diseases based on an individual patient’s
genomic profile.
• Genes, RNAs and proteins are also impacted
by our lifestyle, habits, and environment.
• Wearable devices will be able to record
physiologic data such as temperature, heart
rate, blood pressure, blood oxygenation, heart
rhythm, sleep patterns, and weight.
• Correlational and causal analysis can be
performed.
• Personalized medicine promises to yield more
effective diagnostic measures and treatments
leading to healthier, longer lives and lower
healthcare costs.

Big Data: Genomics and Personalized Medicine
• The amount of data produced by personalized medicine is propelling
healthcare into the realm of big data. Genomic-based medicine offers
tremendous promise and power to revolutionize clinical care, and it
will exponentially change healthcare analytics.
• Genomics produces huge volumes of
data. Each human genome is comprised
of over 3 billion base pairs (1K gigabytes
of data).
• Sequencing human genomes quickly
adds up to hundreds of petabytes of
data; and the data created by-omics
(bioinformatics, proteomics,
metabonomics …) multiplies many times.
• Research and translational medicine
created an analytics discipline called
bioinformatics which developed many
tools and databases used by healthcare
predictive analytics.

HIMSS Big Data Architecture
HIMSS, “A Big Data ReferenceArchitecture”
http://www.himss.org/ResourceLibrary/genResourceFAQ.aspx?ItemNumber=30736

Conceptual Big Data Analytics Architecture
http://hortonworks.com/blog/modern-healthcare-architectures-built-with-hadoop/

Big Data Capabilities
McKinsey – The “Big Data” Revolution in Healthcare http://www.mckinsey.com/insights/health_systems_and_services/the_big-data_revolution_in_us_health_care

Big Data Implementation Success Factors
• Test and learn: An agile approach with rapid releases enables
organizations to fine-tune their projects while they’re in progress. Traditional
legacy systems were better suited to a “waterfall” approach, where
technology was introduced all at once. Big Data projects should focus on
specific business goals and allow cross-pollination of ideas to better
understand what’s possible, making a “rapid release” approach much better.
• Incremental adoption: Build a center of competency and cross-pollinate
expertise among business experts, data scientists, and data engineers. This
approach enables business units to leverage a common talent pool and a
shared approach, eliminating the risk of data silos, providing for common
governance, and avoiding redundant storage and processing by different
departments.
• Change management: Think about key stakeholders for the initiative,
understand their concerns, get their buy in and invest in early pilot systems
that demonstrate the value that can be generated through a Big Data
investment.
http://thinkbiganalytics.com/resources/big-data-whitepaper/right-start-big-data-projects

Working With Unstructured Data:
Text Analytics

Unstructured Data
• Unstructured data has no identifiable structure.
• Typically includes image/objects data (i.e., image, video, & sound
files), text and other data types that are not codified and easily
analyzed using conventional database tools.
• In addition to EMR and other clinical/operational data systems,
consider other sources of untapped knowledge in your healthcare
organization:
– Emails
– Miscellaneous documents (policies, procedures, guidelines)
Source: http://searchstorage.techtarget.com/feature/What-is-unstructured-data-and-how-is-it-different-from-structured-data-in-the-enterprise

Structured and unstructured information
<ICD9>
<Code>413.9</Code>
<Descr>NOS angina
pectoris</Descr>
</ICD9>

Structured versus Unstructured Data
• “The trick is to create value by extracting the right information from
both internal and external data sources. That is what the science of
data and art of business analytics needs to learn to extract from
larger and larger sets of unstructured data.”
http://www.datasciencecentral.com/profiles/blogs/structured-vs-unstructured-
data-the-rise-of-data-anarchy

In Healthcare, 80% Of Data Untapped for KPIs and Analytics
• Needle Stick Injury Rate
• Reintubation Rate
• Ventilator Associated Pneumonia (VAP)
• Blood Stream Infection Due to Central Line
• Urinary Catheter Related Infection
• Overall Employee Satisfaction
• Patient Satisfaction
• Standardized mortality rate (SMR)
• Iatrogenic Pneumothorax
• Decubitus ulcer
• Length of Stay
• ICU readmission rate
• Patients' Fall Rate
• Medication error
• Adverse Events/Error Rate
Key Performance Indicators (KPIs)
Transaction Records Qualitative Human Data Quantitative Machine Data
Admission notes
Discharge summaries
Progress notes
Imaging study results
Consultant reports
Financial
& Operational
Transactions
Medication records
Laboratory results
Physiologic testing
Biometric sensors
RFID tags
Partial Intelligence

The Nuances of Clinical Documentation: Institutional Memory
• Context
 Heart attack mentioned in History of Present Illness
 Heart attack mentioned in Family History
• Meaning (in ED triage notes)
 DOB to indicate Date of Birth
 DOB to indicate Difficulty of Breathing
• Negation
 The patient denied shortness of breath or chest pain

Approaches To Unstructured Information
Text
processing
technologies:
keyword,
NLP,
probabilistic
Taxonomies, ontologies:
• SNOMED
• UMLS,
• ICD9/10
• RxNorm,
• LOINC
Clinical documentation:
• Encounter / admission
/ progress / procedure
notes,
• Discharge summaries,
• Test results
Formal published
medical knowledge:
• literature
(PubMed),
• protocols,
• guidelines
Unconventional sources:
• Social media,
• call center transcripts
• Internet of Things
• Social Media

Healthcare Text Analytics Use Cases
• Clinical chart review is necessary for:
 Compliance Reporting:
 Clinical Quality Measures (CQM’s),
 Core Measures Quarterly,
 PQRS,
 eRx reporting,
 Hospital Inpatient Quality Reporting Program,
 CHIPRA, (Children’s Health Insurance Program Reauthorization Act),
 ACO Programs
 Public Health Data:
 Syndromic surveillance,
 CDC,
 Cancer registries,
 Immunization reports,
 Adverse drug events
 Research and Population Health Management:
 Cohort identification,
 Recruitment of eligible subjects
• Manual chart review is labor intensive, time consuming, and has
human variation

Text Analytics: Understanding Hospital Acquired Conditions
• HAC’s include adverse drug events, hospital acquired infections,
procedural complications and incidents related to falls
 The incidences per 1,000 visits are:
 65 incidents due to adverse drug events
 60 events due to hospital acquired infections
 51 events due to procedural complications
 15 incidents related to falls
• HAI’s result in extra costs of $5-10M /year for average hospital
• ADE’s cost the average hospital about $5.6M/year
• ADE’s increase LOS by 4.6 days and increases costs by at least $5K/incident
Tinoco A, Evans R, et al. J Am Med Inform Assoc 2011;18:491-497
"Computerized surveillance systems that can access both coded and free text data such as that
found in unstructured narratives may improve surveillance without requiring the time and cost
associated with manual chart review, alone.“

Texas Analytics: Looking for Depression and Bipolar Syndromes
• 1 in 10 adults met criteria for current depression, 4.1% met the criteria
for major depression 1
• Median delay from onset of depression to the beginning of treatment
was estimated to be eight years 2
• Treatment prevalence rates of comorbid depression for some chronic
conditions is significantly lower than the expected comorbidity rates
(e.g., 16% treated vs. 45% expected) 3
• Comorbid depression increases medical services costs by average of
$505/member/mo. and has significantly elevated odds ratios with
poorer self care (e.g., high fat intake > 6x/wk, smoking, exercise <
1x/wk) 3
1: MMWR 2010;59(38);1229-1235
2: Melek, S. Halford, M. Measuring the Cost of Undiagnosed Depression. Contingencies. Jul/Aug 2012. Pps. 64-70
3: Melek, S. Norris, D. Chronic conditions and comorbid physchological disorders. Milliman Research. 2008

Texas Analytics Applications in Healthcare
Reconciliation
ID discrepancies between
diagnostic code and clinical notes
Monitor KPIs and Metrics
Reporting
Abstraction
Rapid Chart Access

Natural Language Processing (NLP) for Hawaii Syndromic Surveillance
NLP
Classifier
Triage clerk
Chief Complaint
Cough/Fever
Syndromic Category
• Respiratory
• GI
• Neurological
• Rash
• Hemorrhagic
• Botulinic
• Constitutional
Alarms
Epidemiology Team
Chief Complaint

Managing Structured & Unstructured Data

Healthcare Analytics Maturity Model

Level 8
Cost per Unit of Health Reimbursement & Prescriptive Analytics: Providers Analytic motive expands to wellness management and mass customization of
care. Physicians, hospitals, employers, payers and members/patients collaborate to share risk and reward (e.g., financial reward to patients for
healthy behavior). Analytics expands to include NLP of text, prescriptive analytics, and interventional decision support. Prescriptive analytics are available at
the point of care to improve patient specific outcomes based upon population outcomes. Data content expands to include genomic and familial information.
The EDW is updated within a few minutes of changes in the source systems.
Level 7
Cost per Capita Reimbursement & Predictive Analytics: Analytic motive expands to address diagnosis-based, fixed-fee per capita reimbursement models.
Focus expands from management of cases to collaboration with clinician and payer partners to manage episodes of care, using predictive modeling,
forecasting, and risk stratification to support outreach, triage, escalation and referrals. Patients are flagged in registries who are unable or will not participate
in care protocols. Data content expands to include external pharmacy data and protocol-specific patient reported outcomes. On average, the EDW is updated
within one hour or less of source system changes.
Level 6
Cost per Case Reimbursement & The Triple Aim: The “accountable care organization” shares in the financial risk and reward that is tied to clinical outcomes.
At least 50% of acute care cases are managed under bundled payments. Analytics are available at the point of care to support the Triple Aim of maximizing the
quality of individual patient care, population management, and the economics of care. Data content expands to include bedside devices and detailed activity
based costing. Data governance plays a major role in the accuracy of metrics supporting quality-based compensation plans for clinicians and executives. On
average, the EDW is updated within one day of source system changes. The EDW reports organizationally to a C-level executive who is accountable for
balancing cost of care and quality of care.
Level 5
Clinical Effectiveness & Population Management: Analytic motive is focused on measuring clinical effectiveness that maximizes quality and minimizes waste
and variability. Data governance expands to support care management teams that are focused on improving the health of patient populations. Permanent
multidisciplinary teams are in-place that continuously monitor opportunities to improve quality, and reduce risk and cost, across acute care processes, chronic
diseases, patient safety scenarios, and internal workflows. Precision of registries is improved by including data from lab, pharmacy, and clinical observations in
the definition of the patient cohorts. EDW content is organized into evidence-based, standardized data marts that combine clinical and cost data
associated with patient registries. Data content expands to include insurance claims. On average, the EDW is updated within one week of source system
changes.
Level 4
Automated External Reporting: Analytic motive is focused on consistent, efficient production of reports required for regulatory and accreditation
requirements (e.g. CMS, Joint Commission, tumor registry, communicable diseases); payer incentives (e.g. MU, PQRS, VBP, readmission reduction); and
specialty society databases (e.g. STS,NRMI, Vermont-Oxford). Adherence to industry-standard vocabularies is required. Clinical text data content is available
for simple key word searches. Centralized data governance exists for review and approval of externally released data.
Level 3
Automated Internal Reporting: Analytic motive is focused on consistent, efficient production of reports supporting basic management and operation of the
healthcare organization. Key performance indicators are easily accessible from the executive level to the front-line manager. Corporate and business unit data
analysts meet regularly to collaborate and steer the EDW. Data governance expands to raise the data literacy of the organization and develop a data
acquisition strategy for Levels 4 and above.
Level 2
Standardized Vocabulary & Patient Registries: Master vocabulary and reference data identified and standardized across disparate source system content in
the data warehouse. Naming, definition, and data types are consistent with local standards. Patient registries are defined solely on ICD billing data. Data
governance forms around the definition and evolution of patient registries and master data management.
Level 1
Integrated, Enterprise Data Warehouse: At a minimum, the following data are co-located in a single data warehouse, locally or hosted: HIMSS EMR Stage 3
data, Revenue Cycle, Financial, Costing, Supply Chain, and Patient Experience. Searchable metadata repository is available across the enterprise. Data
content includes insurance claims, if possible. Data warehouse is updated within one month of changes in the source system. Data governance is forming
around the data quality of source systems. The EDW reports organizationally to the CIO.
Level 0
Fragmented Point Solutions: Vendor-based and internally developed applications are used to address specific analytic needs as they arise. The fragmented
Point Solutions are neither co-located in a data warehouse nor otherwise architecturally integrated with one another. Overlapping data content leads to
multiple versions of analytic truth. Reports are labor intensive and inconsistent. Data governance is non-existent.
©
Healthcare Analytic Adoption Model

Healthcare Analytic Adoption Model
Level 8
Cost per Unit of Health Reimbursement &
Prescriptive Analytics
Contracting for & managing health. Customizing
patient care based on population outcomes.
Level 7
Cost per Capita Reimbursement &
Predictive Analytics
Diagnosis-based financial reimbursement,
managing risk proactively, measuring true
outcomes
Level 6
Cost per Case Reimbursement
& The Triple Aim
Procedure-based financial risk and applying
“closed loop” analytics at the point of care
Level 5 Clinical Effectiveness & Population Management Measuring & managing evidence based care
Level 4 Automated External Reporting
Efficient, consistent production; agility, and
governance
Level 3 Automated Internal Reporting
Efficient, consistent production; widespread access
to KPIs
Level 2 Standardized Vocabulary & Patient Registries Relating and organizing the core data
Level 1 Integrated, Enterprise Data Warehouse Foundation of data and technology
Level 0 Fragmented Point Solutions Inefficient, inconsistent versions of the truth
HCAD 6635 Health Information Analytics Copyright © 2016 Frank F. WangCopyright © 2016 Frank F. Wang 72

Population Health Management, Predictive Analytics, Big Data and Text Analytics

Empfohlen

Empfohlen

Weitere ähnliche Inhalte

Was ist angesagt?

Was ist angesagt? (20)

Andere mochten auch

Andere mochten auch (20)

Ähnlich wie Population Health Management, Predictive Analytics, Big Data and Text Analytics

Ähnlich wie Population Health Management, Predictive Analytics, Big Data and Text Analytics (20)

Mehr von Frank Wang

Mehr von Frank Wang (9)

Kürzlich hochgeladen

Kürzlich hochgeladen (20)

Population Health Management, Predictive Analytics, Big Data and Text Analytics